In [ ]:
%%HTML
<script src="require.js"></script>
AI&ML PROJECT - Summer 2023¶
Students :
-Ahmad Khalil Fratekh - 0206882
-Noor Mohammad Albaw - 0201672 Project Overview :¶
The project involves working on a dataset of medical students, working on the dataset we went through multiple operations, which include :
1-Data preparation
2-Classification using sklearn
3-Classification using neural networks from kerasand we have set our target to be the "Diabetes" column
In [ ]:
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import sklearn
from sklearn.impute import SimpleImputer
from scipy import stats
import os
import seaborn as sns
from sklearn.preprocessing import *
from sklearn.model_selection import *
from sklearn.model_selection import cross_val_score
from sklearn.compose import ColumnTransformer
from sklearn.pipeline import Pipeline
from sklearn.preprocessing import StandardScaler
from sklearn.tree import DecisionTreeClassifier
from sklearn.model_selection import GridSearchCV
from sklearn.linear_model import LogisticRegression
from sklearn.ensemble import RandomForestClassifier
from sklearn.neighbors import KNeighborsClassifier
from sklearn.neural_network import MLPClassifier
from sklearn.svm import SVC
import joblib
from sklearn.metrics import *
from sklearn.utils import shuffle
from sklearn import preprocessing
from pandas.plotting import scatter_matrix
import tensorflow as tf
import keras
from keras import layers
import shutil
from tensorflow.keras import initializers
import plotly.express as px
import plotly.io as pio
pio.renderers.default='notebook'
PART 1¶
Data Preparation :
In [ ]:
data = pd.read_csv('medical_students_dataset.csv')
data.shape
Out[ ]:
(200000, 13)
In [ ]:
data.info()
<class 'pandas.core.frame.DataFrame'> RangeIndex: 200000 entries, 0 to 199999 Data columns (total 13 columns): # Column Non-Null Count Dtype --- ------ -------------- ----- 0 Student ID 180000 non-null float64 1 Age 180000 non-null float64 2 Gender 180000 non-null object 3 Height 180000 non-null float64 4 Weight 180000 non-null float64 5 Blood Type 180000 non-null object 6 BMI 180000 non-null float64 7 Temperature 180000 non-null float64 8 Heart Rate 180000 non-null float64 9 Blood Pressure 180000 non-null float64 10 Cholesterol 180000 non-null float64 11 Diabetes 180000 non-null object 12 Smoking 180000 non-null object dtypes: float64(9), object(4) memory usage: 19.8+ MB
In [ ]:
data.isnull().sum()
Out[ ]:
Student ID 20000 Age 20000 Gender 20000 Height 20000 Weight 20000 Blood Type 20000 BMI 20000 Temperature 20000 Heart Rate 20000 Blood Pressure 20000 Cholesterol 20000 Diabetes 20000 Smoking 20000 dtype: int64
Since, as we can see from the plot below, the data is not balanced, we started off by extracting a balanced data to work on¶
In [ ]:
plt.gca().set_title("Diabetes")
sns.countplot(y="Diabetes" , palette='Set3', data = data)
Out[ ]:
<Axes: title={'center': 'Diabetes'}, xlabel='count', ylabel='Diabetes'>
The following three cells of code, implement and visualize the procedure of extracting our balanced dataset :¶
In [ ]:
data_yes = data[data["Diabetes"]=="Yes"]
data_no = data[data["Diabetes"]=="No"]
pd.DataFrame(data_yes)
pd.DataFrame(data_no)
Med_Stu=pd.concat([data_yes[:18000],data_no[:18000]], ignore_index=True)
Med_Stu["Diabetes"]
Out[ ]:
0 Yes
1 Yes
2 Yes
3 Yes
4 Yes
...
35995 No
35996 No
35997 No
35998 No
35999 No
Name: Diabetes, Length: 36000, dtype: object
In [ ]:
print(Med_Stu.shape)
Med_Stu.info()
(36000, 13) <class 'pandas.core.frame.DataFrame'> RangeIndex: 36000 entries, 0 to 35999 Data columns (total 13 columns): # Column Non-Null Count Dtype --- ------ -------------- ----- 0 Student ID 32406 non-null float64 1 Age 32444 non-null float64 2 Gender 32354 non-null object 3 Height 32393 non-null float64 4 Weight 32346 non-null float64 5 Blood Type 32431 non-null object 6 BMI 32382 non-null float64 7 Temperature 32414 non-null float64 8 Heart Rate 32447 non-null float64 9 Blood Pressure 32419 non-null float64 10 Cholesterol 32404 non-null float64 11 Diabetes 36000 non-null object 12 Smoking 32311 non-null object dtypes: float64(9), object(4) memory usage: 3.6+ MB
In [ ]:
plt.gca().set_title("Diabetes")
sns.countplot(y="Diabetes" , palette=['lightsteelblue','slategrey'], data = Med_Stu)
Out[ ]:
<Axes: title={'center': 'Diabetes'}, xlabel='count', ylabel='Diabetes'>
Now that we have extracted our balanced data, it's time to get rid of unnecessary or full of null values column(s), which (is/are) : "Student ID"¶
In [ ]:
Med_Stu_0= Med_Stu.drop("Student ID", axis=1,inplace=True)
In [ ]:
Med_Stu.hist(bins=50, figsize=(20,15), color='dimgray', ec='lavender')
plt.show()
Med_Stu.describe()
Out[ ]:
| Age | Height | Weight | BMI | Temperature | Heart Rate | Blood Pressure | Cholesterol | |
|---|---|---|---|---|---|---|---|---|
| count | 32444.000000 | 32393.000000 | 32346.000000 | 32382.000000 | 32414.000000 | 32447.000000 | 32419.000000 | 32404.000000 |
| mean | 26.043983 | 174.730780 | 70.088455 | 23.392699 | 98.596852 | 79.516627 | 114.530337 | 184.759844 |
| std | 4.894372 | 14.493838 | 17.377104 | 7.093494 | 0.502917 | 11.546663 | 14.398554 | 37.517835 |
| min | 18.000000 | 150.000996 | 40.000578 | 10.081431 | 96.397835 | 60.000000 | 90.000000 | 120.000000 |
| 25% | 22.000000 | 162.149046 | 55.009155 | 17.865348 | 98.260639 | 70.000000 | 102.000000 | 152.000000 |
| 50% | 26.000000 | 174.673148 | 70.082653 | 22.733757 | 98.595614 | 80.000000 | 114.000000 | 185.000000 |
| 75% | 30.000000 | 187.302386 | 85.194960 | 28.084662 | 98.939748 | 90.000000 | 127.000000 | 218.000000 |
| max | 34.000000 | 199.997940 | 99.997668 | 44.314074 | 100.566498 | 99.000000 | 139.000000 | 249.000000 |
In the cell below, we can notice that, there's a couple of missing values in all of the features except for the label which raises the need of using an imputer
In [ ]:
Med_Stu.isnull().sum()
Out[ ]:
Age 3556 Gender 3646 Height 3607 Weight 3654 Blood Type 3569 BMI 3618 Temperature 3586 Heart Rate 3553 Blood Pressure 3581 Cholesterol 3596 Diabetes 0 Smoking 3689 dtype: int64
Before getting to the pipeline, we have to seperate the label from other features, afterwards we have to seperate numercial features from the categorical ones, which is implied in the folloing two cells of code¶
In [ ]:
Med_Stu_1= Med_Stu.drop("Diabetes", axis=1)
Med_Stu_label = Med_Stu["Diabetes"].copy()
Med_Stu_1.head(10)
Out[ ]:
| Age | Gender | Height | Weight | Blood Type | BMI | Temperature | Heart Rate | Blood Pressure | Cholesterol | Smoking | |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 0 | 32.0 | Female | 182.537664 | 55.741083 | A | 16.729017 | 98.260293 | 76.0 | 130.0 | 216.0 | No |
| 1 | 34.0 | Male | NaN | 60.882228 | B | 22.544095 | 98.963569 | 89.0 | 130.0 | 243.0 | NaN |
| 2 | 33.0 | Male | 184.718988 | 93.666944 | NaN | 27.451322 | 98.418213 | 68.0 | 133.0 | 180.0 | Yes |
| 3 | 33.0 | Male | 177.165911 | 68.129149 | O | 21.705642 | 98.201649 | NaN | 116.0 | 143.0 | Yes |
| 4 | 33.0 | Male | 160.463706 | 55.755226 | A | 21.653691 | 99.161461 | 77.0 | NaN | 152.0 | Yes |
| 5 | 34.0 | Male | 174.207898 | NaN | B | 22.292974 | 98.919826 | 74.0 | 129.0 | 139.0 | Yes |
| 6 | 24.0 | Female | 162.044348 | 80.638530 | AB | 30.709647 | 98.996462 | 98.0 | 137.0 | 205.0 | Yes |
| 7 | 21.0 | Male | 171.146689 | 69.056734 | AB | 23.575941 | 98.316070 | NaN | 93.0 | 134.0 | No |
| 8 | 34.0 | Female | 150.942632 | 90.580214 | O | 39.756624 | 97.563234 | 79.0 | 135.0 | 198.0 | No |
| 9 | 21.0 | Male | 159.633475 | 89.877838 | AB | 35.269937 | 98.592047 | 71.0 | 115.0 | 137.0 | No |
In [ ]:
Med_Stu_Cat = Med_Stu_1.select_dtypes(include="object").columns
Med_Stu_Num = Med_Stu_1.select_dtypes(exclude="object").columns
The pipeline:¶
After separating the features from the response: -The cell below shows 2 pipelines that are created, after dividing the dataset into two categories, the first one contains the numeric features, where the second one contains the categorical features.
-The first pipeline, (i.e. "NUM_PIPELINE"), is used to, first, fill the missing values of the numerical features, and then scale the numeric features, to make them easier to compare and to operate on them.
-The second one , (i.e. "CAT_PIPELINE"), is used, first, fill the missing values of the categorical features, and then encode the categorical features using the one-hot encoder, so that we can find the correlation between the label (i.e. "Heart_Disease") and the other non-numeric features in the dataset.
-The last pipeline (i.e. "full_pipeline"), basically, combines the first two together.In [ ]:
NUM_PIPLINE= Pipeline([('num_imp', SimpleImputer(strategy ="median")), ('std_scaler', StandardScaler())])
CAT_PIPLINE= Pipeline([('cat_imp',SimpleImputer(strategy='most_frequent')), ('1_hot_encoder',OneHotEncoder(sparse=False))])
full_pipeline = ColumnTransformer([
("num",NUM_PIPLINE, Med_Stu_Num),
("cat", CAT_PIPLINE,Med_Stu_Cat),
],remainder='passthrough')
med_stu = full_pipeline.fit_transform(Med_Stu_1)
Med_Stu_Prepared=pd.DataFrame(med_stu,columns=full_pipeline.get_feature_names_out())
print(Med_Stu_Prepared.shape)
Med_Stu_Prepared.head()
(36000, 16)
c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\preprocessing\_encoders.py:972: FutureWarning: `sparse` was renamed to `sparse_output` in version 1.2 and will be removed in 1.4. `sparse_output` is ignored unless you leave `sparse` to its default value.
Out[ ]:
| num__Age | num__Height | num__Weight | num__BMI | num__Temperature | num__Heart Rate | num__Blood Pressure | num__Cholesterol | cat__Gender_Female | cat__Gender_Male | cat__Blood Type_A | cat__Blood Type_AB | cat__Blood Type_B | cat__Blood Type_O | cat__Smoking_No | cat__Smoking_Yes | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 0 | 1.282817 | 0.568260 | -0.871013 | -0.980244 | -0.705013 | -0.325128 | 1.135976 | 0.877000 | 1.0 | 0.0 | 1.0 | 0.0 | 0.0 | 0.0 | 1.0 | 0.0 |
| 1 | 1.713266 | -0.003772 | -0.558887 | -0.116245 | 0.768728 | 0.860695 | 1.135976 | 1.635549 | 0.0 | 1.0 | 0.0 | 0.0 | 1.0 | 0.0 | 1.0 | 0.0 |
| 2 | 1.498041 | 0.726921 | 1.431519 | 0.612866 | -0.374085 | -1.054866 | 1.355525 | -0.134399 | 0.0 | 1.0 | 0.0 | 0.0 | 1.0 | 0.0 | 0.0 | 1.0 |
| 3 | 1.498041 | 0.177541 | -0.118916 | -0.240822 | -0.827904 | 0.039740 | 0.111415 | -1.173892 | 0.0 | 1.0 | 0.0 | 0.0 | 0.0 | 1.0 | 0.0 | 1.0 |
| 4 | 1.498041 | -1.037308 | -0.870155 | -0.248541 | 1.183418 | -0.233911 | -0.034951 | -0.921043 | 0.0 | 1.0 | 1.0 | 0.0 | 0.0 | 0.0 | 0.0 | 1.0 |
In [ ]:
Med_Stu_Prepared.hist(bins=50, figsize=(20,15), color='dimgray', ec='midnightblue')
plt.show()
Med_Stu_Prepared.describe()
Out[ ]:
| num__Age | num__Height | num__Weight | num__BMI | num__Temperature | num__Heart Rate | num__Blood Pressure | num__Cholesterol | cat__Gender_Female | cat__Gender_Male | cat__Blood Type_A | cat__Blood Type_AB | cat__Blood Type_B | cat__Blood Type_O | cat__Smoking_No | cat__Smoking_Yes | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| count | 3.600000e+04 | 3.600000e+04 | 3.600000e+04 | 3.600000e+04 | 3.600000e+04 | 3.600000e+04 | 3.600000e+04 | 3.600000e+04 | 36000.000000 | 36000.000000 | 36000.000000 | 36000.000000 | 36000.000000 | 36000.000000 | 36000.000000 | 36000.000000 |
| mean | 2.084752e-16 | 3.042505e-16 | 9.456140e-16 | 2.456800e-16 | -2.901393e-14 | -3.720481e-16 | 3.720481e-16 | 3.821141e-16 | 0.444472 | 0.555528 | 0.221556 | 0.222750 | 0.328278 | 0.227417 | 0.818444 | 0.181556 |
| std | 1.000014e+00 | 1.000014e+00 | 1.000014e+00 | 1.000014e+00 | 1.000014e+00 | 1.000014e+00 | 1.000014e+00 | 1.000014e+00 | 0.496914 | 0.496914 | 0.415299 | 0.416098 | 0.469593 | 0.419170 | 0.385483 | 0.385483 |
| min | -1.730328e+00 | -1.798322e+00 | -1.826641e+00 | -1.967937e+00 | -4.607862e+00 | -1.784603e+00 | -1.791342e+00 | -1.820064e+00 | 0.000000 | 0.000000 | 0.000000 | 0.000000 | 0.000000 | 0.000000 | 0.000000 | 0.000000 |
| 25% | -8.694298e-01 | -8.102296e-01 | -8.100104e-01 | -7.342704e-01 | -6.183490e-01 | -7.812143e-01 | -7.667805e-01 | -8.086649e-01 | 0.000000 | 0.000000 | 0.000000 | 0.000000 | 0.000000 | 0.000000 | 1.000000 | 0.000000 |
| 50% | -8.531266e-03 | -3.771884e-03 | -3.164960e-04 | -8.806557e-02 | -2.335110e-03 | 3.974028e-02 | -3.495096e-02 | 6.073076e-03 | 0.000000 | 1.000000 | 0.000000 | 0.000000 | 0.000000 | 0.000000 | 1.000000 | 0.000000 |
| 75% | 8.523673e-01 | 8.092148e-01 | 8.156271e-01 | 5.953376e-01 | 6.297012e-01 | 7.694777e-01 | 8.432445e-01 | 8.208111e-01 | 1.000000 | 1.000000 | 0.000000 | 0.000000 | 1.000000 | 0.000000 | 1.000000 | 0.000000 |
| max | 1.713266e+00 | 1.838249e+00 | 1.815866e+00 | 3.118318e+00 | 4.127724e+00 | 1.772867e+00 | 1.794623e+00 | 1.804116e+00 | 1.000000 | 1.000000 | 1.000000 | 1.000000 | 1.000000 | 1.000000 | 1.000000 | 1.000000 |
The cell below shows the re-appending of the label to the dataset after it is prepared
In [ ]:
Med_Stu_Prepared["Diabetes"] = Med_Stu_label
Med_Stu_Prepared.shape
Out[ ]:
(36000, 17)
Encoding of the label using the label encoder
In [ ]:
label_encoder = preprocessing.LabelEncoder()
Med_Stu_Prepared["Diabetes"]= label_encoder.fit_transform(Med_Stu_Prepared["Diabetes"])
Med_Stu_Prepared=shuffle(Med_Stu_Prepared)
Med_Stu_Prepared.head()
Out[ ]:
| num__Age | num__Height | num__Weight | num__BMI | num__Temperature | num__Heart Rate | num__Blood Pressure | num__Cholesterol | cat__Gender_Female | cat__Gender_Male | cat__Blood Type_A | cat__Blood Type_AB | cat__Blood Type_B | cat__Blood Type_O | cat__Smoking_No | cat__Smoking_Yes | Diabetes | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 2311 | 0.852367 | 0.402548 | -0.214484 | -0.422540 | 1.147982 | 0.587043 | -0.620415 | 0.427489 | 0.0 | 1.0 | 0.0 | 0.0 | 1.0 | 0.0 | 1.0 | 0.0 | 1 |
| 21800 | -0.223756 | 0.645932 | -0.000316 | -0.052685 | 2.355241 | 0.313392 | 1.794623 | -0.724382 | 0.0 | 1.0 | 1.0 | 0.0 | 0.0 | 0.0 | 1.0 | 0.0 | 0 |
| 15970 | -0.008531 | 1.061236 | -0.847606 | -1.139046 | -0.836823 | -0.233911 | 1.794623 | 1.157944 | 1.0 | 0.0 | 0.0 | 1.0 | 0.0 | 0.0 | 0.0 | 1.0 | 1 |
| 18948 | -1.730328 | -0.606676 | -1.044023 | -0.627134 | -1.353082 | 0.860695 | 0.330964 | 1.466983 | 1.0 | 0.0 | 0.0 | 1.0 | 0.0 | 0.0 | 1.0 | 0.0 | 0 |
| 22014 | 0.637143 | 1.036016 | -0.008434 | -0.555379 | -0.335294 | 1.590432 | -1.205878 | 1.242227 | 0.0 | 1.0 | 0.0 | 0.0 | 1.0 | 0.0 | 1.0 | 0.0 | 0 |
The cell below finds the correlation between the label (i.e. "Diabetes"), and the remaining features in the dataset
In [ ]:
corr_matrix=Med_Stu_Prepared.corr(numeric_only=True)
corr_matrix["Diabetes"].sort_values(ascending=False)
Out[ ]:
Diabetes 1.000000 num__Age 0.007611 num__Heart Rate 0.005605 cat__Smoking_No 0.004756 cat__Blood Type_B 0.003668 cat__Blood Type_A 0.003612 num__Weight 0.002761 num__Cholesterol 0.001233 cat__Gender_Female 0.000503 num__BMI 0.000398 cat__Gender_Male -0.000503 num__Blood Pressure -0.002450 cat__Blood Type_O -0.003777 cat__Blood Type_AB -0.003939 cat__Smoking_Yes -0.004756 num__Height -0.005167 num__Temperature -0.010606 Name: Diabetes, dtype: float64
In [ ]:
attributes = ["num__BMI", "num__Weight","num__Height"]
fig = px.scatter_matrix(Med_Stu_Prepared,
dimensions=attributes,
color="Diabetes",
title="Scatter matrix of medical students set",
labels={col:col.replace('num__', ' ') for col in Med_Stu_Prepared.columns})
fig.show()
In [ ]:
ax=sns.set(rc= {"figure.figsize": (8,4)})
ax=sns.boxplot(
data=Med_Stu_Prepared, x=Med_Stu_Prepared["num__BMI"].round(2), y=Med_Stu_Prepared["num__Weight"].round(0),
notch=True, showcaps=False,
flierprops={"marker": "x"},
boxprops={"facecolor": (.4, .6, .8, .5)},
medianprops={"color": "coral"},orient="h",dodge=True
)
ax.set(ylabel="Weight")
ax.set(xlabel="BMI")
Out[ ]:
[Text(0.5, 0, 'BMI')]
In [ ]:
x1=(Med_Stu_Prepared[(Med_Stu_Prepared['Diabetes']==1.0) & (Med_Stu_Prepared['cat__Blood Type_B']==1.0)])
a=x1.shape[0]
x2=(Med_Stu_Prepared[(Med_Stu_Prepared['Diabetes']==1.0) & (Med_Stu_Prepared['cat__Blood Type_A']==1.0)])
b=x2.shape[0]
x3=(Med_Stu_Prepared[(Med_Stu_Prepared['Diabetes']==1.0) & (Med_Stu_Prepared['cat__Blood Type_O']==1.0)])
c=x3.shape[0]
x4=(Med_Stu_Prepared[(Med_Stu_Prepared['Diabetes']==1.0) & (Med_Stu_Prepared['cat__Blood Type_AB']==1.0)])
d=x4.shape[0]
list=[a,b,c,d]
colors=['coral','lightcoral','steelblue','aliceblue']
patterns=('*','\\','+','|')
labels=['B','A','O','AB']
bar=plt.bar(labels,list,label='Blood Type', width=0.35, color=colors)
for i,x in zip(bar,patterns):
i.set_hatch(x)
plt.ylabel("Diabetes ")
plt.xlabel('Blood Categories')
plt.title('Relation Between blood type and Diabetes ')
plt.show()
The following three cells of code demonstrate the procedure of splitting the data into a training set and a test set with no label¶
In [ ]:
import numpy as np
def split_train_test (data, test_ratio):
shuffled_indices = np.random.permutation(len(data))
test_set_size = int(len(data)*test_ratio)
test_indices = shuffled_indices[:test_set_size]
train_indices = shuffled_indices[test_set_size:]
return data.iloc[train_indices], data.iloc[test_indices]
train_set, test_set = split_train_test(Med_Stu_Prepared, 0.2)
print (len(train_set),"train +", len(test_set), "test")
28800 train + 7200 test
In [ ]:
train = train_set.drop("Diabetes", axis=1)
l1 = train_set["Diabetes"].copy()
len(l1)
Out[ ]:
28800
In [ ]:
test = test_set.drop("Diabetes", axis=1)
l2 = test_set["Diabetes"].copy()
test.shape
Out[ ]:
(7200, 16)
PART 2¶
Classifiers used:
1-DecisionTreeClassifier
2-LogisticRegression
3-SVC
4-RandomForestClassifier
5-KNeighborsClassifier
6-MLPClassifier
7-CalibratedClassifierCVThe model is trained and the accuracy is found on and for each one of the classifiers mentioned above
using the grid search, different parameters are tested for each one of the classifiers
1-DecisionTreeClassifier¶
In [ ]:
tree_clf = DecisionTreeClassifier(random_state=42)
tree_params = [
{'criterion': ['gini', 'entropy', 'log_loss'],
'splitter': ['best','random'],
'max_depth': [3,4,5]}]
grid_search = GridSearchCV(tree_clf,tree_params, cv=3,
scoring='accuracy')
grid_search.fit(train, l1)
Out[ ]:
GridSearchCV(cv=3, estimator=DecisionTreeClassifier(random_state=42),
param_grid=[{'criterion': ['gini', 'entropy', 'log_loss'],
'max_depth': [3, 4, 5],
'splitter': ['best', 'random']}],
scoring='accuracy')In a Jupyter environment, please rerun this cell to show the HTML representation or trust the notebook. On GitHub, the HTML representation is unable to render, please try loading this page with nbviewer.org.
GridSearchCV(cv=3, estimator=DecisionTreeClassifier(random_state=42),
param_grid=[{'criterion': ['gini', 'entropy', 'log_loss'],
'max_depth': [3, 4, 5],
'splitter': ['best', 'random']}],
scoring='accuracy')DecisionTreeClassifier(random_state=42)
DecisionTreeClassifier(random_state=42)
The accuracy, precision, recall, and f1 scores, for the test set :
In [ ]:
P1=grid_search.predict(train)
P2=grid_search.predict(test)
print("Best parameter = "+str(grid_search.best_params_))
print("Best estimator = "+str(grid_search.best_estimator_))
print('The accuracy for DTC = '+str(grid_search.best_score_))
print("precision = "+str(precision_score(l2,P2)))
print("recall = "+str(recall_score(l2,P2)))
print("f1_score = "+str(f1_score(l2,P2)))
Best parameter = {'criterion': 'entropy', 'max_depth': 5, 'splitter': 'best'}
Best estimator = DecisionTreeClassifier(criterion='entropy', max_depth=5, random_state=42)
The accuracy for DTC = 0.5084722222222222
precision = 0.5045632333767927
recall = 0.21614074280927115
f1_score = 0.3026392961876832
The confusion matrix for:
1-training set
2-testing setIn [ ]:
ConfusionMatrixDisplay.from_predictions(l1,P1)
ConfusionMatrixDisplay.from_predictions(l2,P2)
plt.show()
2-LogisticRegression¶
In [ ]:
log_reg = LogisticRegression(max_iter=1000)
log_params = [
{'C': [1,10,100],
'fit_intercept': [True,False],
'n_jobs': [-1,10,16],
}]
grid_search = GridSearchCV(log_reg,log_params, cv=3,
scoring='accuracy')
grid_search.fit(train, l1)
Out[ ]:
GridSearchCV(cv=3, estimator=LogisticRegression(max_iter=1000),
param_grid=[{'C': [1, 10, 100], 'fit_intercept': [True, False],
'n_jobs': [-1, 10, 16]}],
scoring='accuracy')In a Jupyter environment, please rerun this cell to show the HTML representation or trust the notebook. On GitHub, the HTML representation is unable to render, please try loading this page with nbviewer.org.
GridSearchCV(cv=3, estimator=LogisticRegression(max_iter=1000),
param_grid=[{'C': [1, 10, 100], 'fit_intercept': [True, False],
'n_jobs': [-1, 10, 16]}],
scoring='accuracy')LogisticRegression(max_iter=1000)
LogisticRegression(max_iter=1000)
The accuracy, precision, recall, and f1 scores, for the test set :
In [ ]:
P1=grid_search.predict(train)
P2=grid_search.predict(test)
print("Best parameter = "+str(grid_search.best_params_))
print("Best estimator = "+str(grid_search.best_estimator_))
print('The accuracy for LogisticRegression = '+str(grid_search.best_score_))
print("precision = "+str(precision_score(l2,P2)))
print("recall = "+str(recall_score(l2,P2)))
print("f1_score = "+str(f1_score(l2,P2)))
Best parameter = {'C': 1, 'fit_intercept': True, 'n_jobs': -1}
Best estimator = LogisticRegression(C=1, max_iter=1000, n_jobs=-1)
The accuracy for LogisticRegression = 0.5050347222222222
precision = 0.5020544427324088
recall = 0.5459368891371126
f1_score = 0.5230769230769231
The confusion matrix for:
1-training set
2-testing setIn [ ]:
ConfusionMatrixDisplay.from_predictions(l1,P1)
ConfusionMatrixDisplay.from_predictions(l2,P2)
plt.show()
3-SupportVectorClassifier¶
In [ ]:
svm = SVC()
param_grid = [
{
'gamma': [2, 1],
'C':[0.001,0.1],
}]
grid_search = GridSearchCV(svm, param_grid, cv=3,scoring='accuracy')
grid_search.fit(train[:7000],l1[:7000])
Out[ ]:
GridSearchCV(cv=3, estimator=SVC(),
param_grid=[{'C': [0.001, 0.1], 'gamma': [2, 1]}],
scoring='accuracy')In a Jupyter environment, please rerun this cell to show the HTML representation or trust the notebook. On GitHub, the HTML representation is unable to render, please try loading this page with nbviewer.org.
GridSearchCV(cv=3, estimator=SVC(),
param_grid=[{'C': [0.001, 0.1], 'gamma': [2, 1]}],
scoring='accuracy')SVC()
SVC()
The accuracy, precision, recall, and f1 scores, for the test set :
In [ ]:
P1=grid_search.predict(train)
P2=grid_search.predict(test)
print("Best parameter = "+str(grid_search.best_params_))
print("Best estimator = "+str(grid_search.best_estimator_))
print('The accuracy for SVC = '+str(grid_search.best_score_))
print("precision = "+str(precision_score(l2,P2)))
print("recall = "+str(recall_score(l2,P2)))
print("f1_score = "+str(f1_score(l2,P2)))
Best parameter = {'C': 0.001, 'gamma': 2}
Best estimator = SVC(C=0.001, gamma=2)
The accuracy for SVC = 0.5018571633381829
precision = 0.4973611111111111
recall = 1.0
f1_score = 0.6643168537241443
The confusion matrix for:
1-training set
2-testing setIn [ ]:
ConfusionMatrixDisplay.from_predictions(l1,P1)
ConfusionMatrixDisplay.from_predictions(l2,P2)
plt.show()
4-RandomForestClassifier¶
In [ ]:
RFC= RandomForestClassifier()
param_grid = [{
'n_estimators':[500,1000],
'min_samples_split':[2,3],
'max_depth' :[2,6]
}]
grid_search = GridSearchCV(RFC, param_grid, cv=3, scoring='accuracy',error_score='raise')
grid_search.fit(train[:15000],l1[:15000])
Out[ ]:
GridSearchCV(cv=3, error_score='raise', estimator=RandomForestClassifier(),
param_grid=[{'max_depth': [2, 6], 'min_samples_split': [2, 3],
'n_estimators': [500, 1000]}],
scoring='accuracy')In a Jupyter environment, please rerun this cell to show the HTML representation or trust the notebook. On GitHub, the HTML representation is unable to render, please try loading this page with nbviewer.org.
GridSearchCV(cv=3, error_score='raise', estimator=RandomForestClassifier(),
param_grid=[{'max_depth': [2, 6], 'min_samples_split': [2, 3],
'n_estimators': [500, 1000]}],
scoring='accuracy')RandomForestClassifier()
RandomForestClassifier()
The accuracy, precision, recall, and f1 scores, for the test set :
In [ ]:
P1=grid_search.predict(train)
P2=grid_search.predict(test)
print("Best parameter = "+str(grid_search.best_params_))
print("Best estimator = "+str(grid_search.best_estimator_))
print('The accuracy for RFC = '+str(grid_search.best_score_))
print("precision = "+str(precision_score(l2,P2)))
print("recall = "+str(recall_score(l2,P2)))
print("f1_score = "+str(f1_score(l2,P2)))
Best parameter = {'max_depth': 6, 'min_samples_split': 2, 'n_estimators': 1000}
Best estimator = RandomForestClassifier(max_depth=6, n_estimators=1000)
The accuracy for RFC = 0.5142666666666666
precision = 0.5191424196018377
recall = 0.3786651773247696
f1_score = 0.4379137736153722
The confusion matrix for:
1-training set
2-testing setIn [ ]:
ConfusionMatrixDisplay.from_predictions(l1,P1)
ConfusionMatrixDisplay.from_predictions(l2,P2)
plt.show()
5-KNeighborsClassifier¶
In [ ]:
knn_clf = KNeighborsClassifier()
param_grid = [{
'n_neighbors':[3,4,5],
'weights':['uniform','distance']
}]
grid_search = GridSearchCV(knn_clf, param_grid, cv=3,scoring='accuracy')
grid_search.fit(train,l1)
c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\model_selection\_validation.py:824: UserWarning:
Scoring failed. The score on this train-test partition for these parameters will be set to nan. Details:
Traceback (most recent call last):
File "c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\model_selection\_validation.py", line 813, in _score
scores = scorer(estimator, X_test, y_test)
File "c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\metrics\_scorer.py", line 266, in __call__
return self._score(partial(_cached_call, None), estimator, X, y_true, **_kwargs)
File "c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\metrics\_scorer.py", line 353, in _score
y_pred = method_caller(estimator, "predict", X)
File "c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\metrics\_scorer.py", line 86, in _cached_call
result, _ = _get_response_values(
File "c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\utils\_response.py", line 85, in _get_response_values
y_pred = prediction_method(X)
File "c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\neighbors\_classification.py", line 246, in predict
if self._fit_method == "brute" and ArgKminClassMode.is_usable_for(
File "c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\metrics\_pairwise_distances_reduction\_dispatcher.py", line 471, in is_usable_for
ArgKmin.is_usable_for(X, Y, metric)
File "c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\metrics\_pairwise_distances_reduction\_dispatcher.py", line 115, in is_usable_for
and (is_numpy_c_ordered(X) or is_valid_sparse_matrix(X))
File "c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\metrics\_pairwise_distances_reduction\_dispatcher.py", line 99, in is_numpy_c_ordered
return hasattr(X, "flags") and X.flags.c_contiguous
AttributeError: 'Flags' object has no attribute 'c_contiguous'
c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\model_selection\_validation.py:824: UserWarning:
Scoring failed. The score on this train-test partition for these parameters will be set to nan. Details:
Traceback (most recent call last):
File "c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\model_selection\_validation.py", line 813, in _score
scores = scorer(estimator, X_test, y_test)
File "c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\metrics\_scorer.py", line 266, in __call__
return self._score(partial(_cached_call, None), estimator, X, y_true, **_kwargs)
File "c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\metrics\_scorer.py", line 353, in _score
y_pred = method_caller(estimator, "predict", X)
File "c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\metrics\_scorer.py", line 86, in _cached_call
result, _ = _get_response_values(
File "c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\utils\_response.py", line 85, in _get_response_values
y_pred = prediction_method(X)
File "c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\neighbors\_classification.py", line 246, in predict
if self._fit_method == "brute" and ArgKminClassMode.is_usable_for(
File "c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\metrics\_pairwise_distances_reduction\_dispatcher.py", line 471, in is_usable_for
ArgKmin.is_usable_for(X, Y, metric)
File "c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\metrics\_pairwise_distances_reduction\_dispatcher.py", line 115, in is_usable_for
and (is_numpy_c_ordered(X) or is_valid_sparse_matrix(X))
File "c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\metrics\_pairwise_distances_reduction\_dispatcher.py", line 99, in is_numpy_c_ordered
return hasattr(X, "flags") and X.flags.c_contiguous
AttributeError: 'Flags' object has no attribute 'c_contiguous'
c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\model_selection\_validation.py:824: UserWarning:
Scoring failed. The score on this train-test partition for these parameters will be set to nan. Details:
Traceback (most recent call last):
File "c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\model_selection\_validation.py", line 813, in _score
scores = scorer(estimator, X_test, y_test)
File "c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\metrics\_scorer.py", line 266, in __call__
return self._score(partial(_cached_call, None), estimator, X, y_true, **_kwargs)
File "c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\metrics\_scorer.py", line 353, in _score
y_pred = method_caller(estimator, "predict", X)
File "c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\metrics\_scorer.py", line 86, in _cached_call
result, _ = _get_response_values(
File "c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\utils\_response.py", line 85, in _get_response_values
y_pred = prediction_method(X)
File "c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\neighbors\_classification.py", line 246, in predict
if self._fit_method == "brute" and ArgKminClassMode.is_usable_for(
File "c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\metrics\_pairwise_distances_reduction\_dispatcher.py", line 471, in is_usable_for
ArgKmin.is_usable_for(X, Y, metric)
File "c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\metrics\_pairwise_distances_reduction\_dispatcher.py", line 115, in is_usable_for
and (is_numpy_c_ordered(X) or is_valid_sparse_matrix(X))
File "c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\metrics\_pairwise_distances_reduction\_dispatcher.py", line 99, in is_numpy_c_ordered
return hasattr(X, "flags") and X.flags.c_contiguous
AttributeError: 'Flags' object has no attribute 'c_contiguous'
c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\model_selection\_validation.py:824: UserWarning:
Scoring failed. The score on this train-test partition for these parameters will be set to nan. Details:
Traceback (most recent call last):
File "c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\model_selection\_validation.py", line 813, in _score
scores = scorer(estimator, X_test, y_test)
File "c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\metrics\_scorer.py", line 266, in __call__
return self._score(partial(_cached_call, None), estimator, X, y_true, **_kwargs)
File "c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\metrics\_scorer.py", line 353, in _score
y_pred = method_caller(estimator, "predict", X)
File "c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\metrics\_scorer.py", line 86, in _cached_call
result, _ = _get_response_values(
File "c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\utils\_response.py", line 85, in _get_response_values
y_pred = prediction_method(X)
File "c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\neighbors\_classification.py", line 246, in predict
if self._fit_method == "brute" and ArgKminClassMode.is_usable_for(
File "c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\metrics\_pairwise_distances_reduction\_dispatcher.py", line 471, in is_usable_for
ArgKmin.is_usable_for(X, Y, metric)
File "c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\metrics\_pairwise_distances_reduction\_dispatcher.py", line 115, in is_usable_for
and (is_numpy_c_ordered(X) or is_valid_sparse_matrix(X))
File "c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\metrics\_pairwise_distances_reduction\_dispatcher.py", line 99, in is_numpy_c_ordered
return hasattr(X, "flags") and X.flags.c_contiguous
AttributeError: 'Flags' object has no attribute 'c_contiguous'
c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\model_selection\_validation.py:824: UserWarning:
Scoring failed. The score on this train-test partition for these parameters will be set to nan. Details:
Traceback (most recent call last):
File "c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\model_selection\_validation.py", line 813, in _score
scores = scorer(estimator, X_test, y_test)
File "c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\metrics\_scorer.py", line 266, in __call__
return self._score(partial(_cached_call, None), estimator, X, y_true, **_kwargs)
File "c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\metrics\_scorer.py", line 353, in _score
y_pred = method_caller(estimator, "predict", X)
File "c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\metrics\_scorer.py", line 86, in _cached_call
result, _ = _get_response_values(
File "c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\utils\_response.py", line 85, in _get_response_values
y_pred = prediction_method(X)
File "c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\neighbors\_classification.py", line 246, in predict
if self._fit_method == "brute" and ArgKminClassMode.is_usable_for(
File "c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\metrics\_pairwise_distances_reduction\_dispatcher.py", line 471, in is_usable_for
ArgKmin.is_usable_for(X, Y, metric)
File "c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\metrics\_pairwise_distances_reduction\_dispatcher.py", line 115, in is_usable_for
and (is_numpy_c_ordered(X) or is_valid_sparse_matrix(X))
File "c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\metrics\_pairwise_distances_reduction\_dispatcher.py", line 99, in is_numpy_c_ordered
return hasattr(X, "flags") and X.flags.c_contiguous
AttributeError: 'Flags' object has no attribute 'c_contiguous'
c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\model_selection\_validation.py:824: UserWarning:
Scoring failed. The score on this train-test partition for these parameters will be set to nan. Details:
Traceback (most recent call last):
File "c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\model_selection\_validation.py", line 813, in _score
scores = scorer(estimator, X_test, y_test)
File "c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\metrics\_scorer.py", line 266, in __call__
return self._score(partial(_cached_call, None), estimator, X, y_true, **_kwargs)
File "c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\metrics\_scorer.py", line 353, in _score
y_pred = method_caller(estimator, "predict", X)
File "c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\metrics\_scorer.py", line 86, in _cached_call
result, _ = _get_response_values(
File "c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\utils\_response.py", line 85, in _get_response_values
y_pred = prediction_method(X)
File "c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\neighbors\_classification.py", line 246, in predict
if self._fit_method == "brute" and ArgKminClassMode.is_usable_for(
File "c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\metrics\_pairwise_distances_reduction\_dispatcher.py", line 471, in is_usable_for
ArgKmin.is_usable_for(X, Y, metric)
File "c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\metrics\_pairwise_distances_reduction\_dispatcher.py", line 115, in is_usable_for
and (is_numpy_c_ordered(X) or is_valid_sparse_matrix(X))
File "c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\metrics\_pairwise_distances_reduction\_dispatcher.py", line 99, in is_numpy_c_ordered
return hasattr(X, "flags") and X.flags.c_contiguous
AttributeError: 'Flags' object has no attribute 'c_contiguous'
c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\model_selection\_validation.py:824: UserWarning:
Scoring failed. The score on this train-test partition for these parameters will be set to nan. Details:
Traceback (most recent call last):
File "c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\model_selection\_validation.py", line 813, in _score
scores = scorer(estimator, X_test, y_test)
File "c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\metrics\_scorer.py", line 266, in __call__
return self._score(partial(_cached_call, None), estimator, X, y_true, **_kwargs)
File "c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\metrics\_scorer.py", line 353, in _score
y_pred = method_caller(estimator, "predict", X)
File "c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\metrics\_scorer.py", line 86, in _cached_call
result, _ = _get_response_values(
File "c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\utils\_response.py", line 85, in _get_response_values
y_pred = prediction_method(X)
File "c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\neighbors\_classification.py", line 246, in predict
if self._fit_method == "brute" and ArgKminClassMode.is_usable_for(
File "c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\metrics\_pairwise_distances_reduction\_dispatcher.py", line 471, in is_usable_for
ArgKmin.is_usable_for(X, Y, metric)
File "c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\metrics\_pairwise_distances_reduction\_dispatcher.py", line 115, in is_usable_for
and (is_numpy_c_ordered(X) or is_valid_sparse_matrix(X))
File "c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\metrics\_pairwise_distances_reduction\_dispatcher.py", line 99, in is_numpy_c_ordered
return hasattr(X, "flags") and X.flags.c_contiguous
AttributeError: 'Flags' object has no attribute 'c_contiguous'
c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\model_selection\_validation.py:824: UserWarning:
Scoring failed. The score on this train-test partition for these parameters will be set to nan. Details:
Traceback (most recent call last):
File "c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\model_selection\_validation.py", line 813, in _score
scores = scorer(estimator, X_test, y_test)
File "c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\metrics\_scorer.py", line 266, in __call__
return self._score(partial(_cached_call, None), estimator, X, y_true, **_kwargs)
File "c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\metrics\_scorer.py", line 353, in _score
y_pred = method_caller(estimator, "predict", X)
File "c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\metrics\_scorer.py", line 86, in _cached_call
result, _ = _get_response_values(
File "c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\utils\_response.py", line 85, in _get_response_values
y_pred = prediction_method(X)
File "c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\neighbors\_classification.py", line 246, in predict
if self._fit_method == "brute" and ArgKminClassMode.is_usable_for(
File "c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\metrics\_pairwise_distances_reduction\_dispatcher.py", line 471, in is_usable_for
ArgKmin.is_usable_for(X, Y, metric)
File "c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\metrics\_pairwise_distances_reduction\_dispatcher.py", line 115, in is_usable_for
and (is_numpy_c_ordered(X) or is_valid_sparse_matrix(X))
File "c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\metrics\_pairwise_distances_reduction\_dispatcher.py", line 99, in is_numpy_c_ordered
return hasattr(X, "flags") and X.flags.c_contiguous
AttributeError: 'Flags' object has no attribute 'c_contiguous'
c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\model_selection\_validation.py:824: UserWarning:
Scoring failed. The score on this train-test partition for these parameters will be set to nan. Details:
Traceback (most recent call last):
File "c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\model_selection\_validation.py", line 813, in _score
scores = scorer(estimator, X_test, y_test)
File "c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\metrics\_scorer.py", line 266, in __call__
return self._score(partial(_cached_call, None), estimator, X, y_true, **_kwargs)
File "c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\metrics\_scorer.py", line 353, in _score
y_pred = method_caller(estimator, "predict", X)
File "c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\metrics\_scorer.py", line 86, in _cached_call
result, _ = _get_response_values(
File "c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\utils\_response.py", line 85, in _get_response_values
y_pred = prediction_method(X)
File "c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\neighbors\_classification.py", line 246, in predict
if self._fit_method == "brute" and ArgKminClassMode.is_usable_for(
File "c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\metrics\_pairwise_distances_reduction\_dispatcher.py", line 471, in is_usable_for
ArgKmin.is_usable_for(X, Y, metric)
File "c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\metrics\_pairwise_distances_reduction\_dispatcher.py", line 115, in is_usable_for
and (is_numpy_c_ordered(X) or is_valid_sparse_matrix(X))
File "c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\metrics\_pairwise_distances_reduction\_dispatcher.py", line 99, in is_numpy_c_ordered
return hasattr(X, "flags") and X.flags.c_contiguous
AttributeError: 'Flags' object has no attribute 'c_contiguous'
c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\model_selection\_search.py:976: UserWarning:
One or more of the test scores are non-finite: [ nan 0.55399306 nan 0.55614583 nan 0.55208333]
Out[ ]:
GridSearchCV(cv=3, estimator=KNeighborsClassifier(),
param_grid=[{'n_neighbors': [3, 4, 5],
'weights': ['uniform', 'distance']}],
scoring='accuracy')In a Jupyter environment, please rerun this cell to show the HTML representation or trust the notebook. On GitHub, the HTML representation is unable to render, please try loading this page with nbviewer.org.
GridSearchCV(cv=3, estimator=KNeighborsClassifier(),
param_grid=[{'n_neighbors': [3, 4, 5],
'weights': ['uniform', 'distance']}],
scoring='accuracy')KNeighborsClassifier()
KNeighborsClassifier()
The accuracy, precision, recall, and f1 scores, for the test set :
In [ ]:
P1=grid_search.predict(train)
P2=grid_search.predict(test)
print("Best parameter = "+str(grid_search.best_params_))
print("Best estimator = "+str(grid_search.best_estimator_))
print('The accuracy for KNeighborsClassifier = '+str(grid_search.best_score_))
print("precision = "+str(precision_score(l2,P2)))
print("recall = "+str(recall_score(l2,P2)))
print("f1_score = "+str(f1_score(l2,P2)))
Best parameter = {'n_neighbors': 4, 'weights': 'distance'}
Best estimator = KNeighborsClassifier(n_neighbors=4, weights='distance')
The accuracy for KNeighborsClassifier = 0.5561458333333333
precision = 0.5614475998994722
recall = 0.623848087126501
f1_score = 0.591005291005291
The confusion matrix for:
1-training set
2-testing setIn [ ]:
ConfusionMatrixDisplay.from_predictions(l1,P1)
ConfusionMatrixDisplay.from_predictions(l2,P2)
plt.show()
6-MLPClassifier¶
In [ ]:
mlp_clf = MLPClassifier(max_iter = 2000)
param_grid = [{
'activation':['identity', 'logistic'],
'alpha':[0.0001,0.001],
'learning_rate' :['constant','invscaling']
}]
grid_search = GridSearchCV(mlp_clf, param_grid, cv=3,scoring='accuracy')
grid_search.fit(train,l1)
Out[ ]:
GridSearchCV(cv=3, estimator=MLPClassifier(max_iter=2000),
param_grid=[{'activation': ['identity', 'logistic'],
'alpha': [0.0001, 0.001],
'learning_rate': ['constant', 'invscaling']}],
scoring='accuracy')In a Jupyter environment, please rerun this cell to show the HTML representation or trust the notebook. On GitHub, the HTML representation is unable to render, please try loading this page with nbviewer.org.
GridSearchCV(cv=3, estimator=MLPClassifier(max_iter=2000),
param_grid=[{'activation': ['identity', 'logistic'],
'alpha': [0.0001, 0.001],
'learning_rate': ['constant', 'invscaling']}],
scoring='accuracy')MLPClassifier(max_iter=2000)
MLPClassifier(max_iter=2000)
The accuracy, precision, recall, and f1 scores, for the test set :
In [ ]:
P1=grid_search.predict(train)
P2=grid_search.predict(test)
print("Best parameter = "+str(grid_search.best_params_))
print("Best estimator = "+str(grid_search.best_estimator_))
print('The accuracy for MLPClassifier = '+str(grid_search.best_score_))
print("precision = "+str(precision_score(l2,P2)))
print("recall = "+str(recall_score(l2,P2)))
print("f1_score = "+str(f1_score(l2,P2)))
Best parameter = {'activation': 'logistic', 'alpha': 0.0001, 'learning_rate': 'invscaling'}
Best estimator = MLPClassifier(activation='logistic', learning_rate='invscaling', max_iter=2000)
The accuracy for MLPClassifier = 0.5057986111111111
precision = 0.5069637883008357
recall = 0.6607092990784696
f1_score = 0.5737148399612026
The confusion matrix for:
1-training set
2-testing setIn [ ]:
ConfusionMatrixDisplay.from_predictions(l1,P1)
ConfusionMatrixDisplay.from_predictions(l2,P2)
plt.show()
7-CalibratedClassifierCV¶
In [ ]:
from sklearn.calibration import CalibratedClassifierCV
CC = CalibratedClassifierCV()
param_grid = [{
'method' :['sigmoid', 'isotonic'],
'estimator' : ['estimator instance', None],
'n_jobs': [-1,10,16]
}]
grid_search = GridSearchCV(CC, param_grid, cv=3,scoring='accuracy')
grid_search.fit(train,l1)
c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\model_selection\_validation.py:425: FitFailedWarning:
18 fits failed out of a total of 36.
The score on these train-test partitions for these parameters will be set to nan.
If these failures are not expected, you can try to debug them by setting error_score='raise'.
Below are more details about the failures:
--------------------------------------------------------------------------------
18 fits failed with the following error:
Traceback (most recent call last):
File "c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\model_selection\_validation.py", line 732, in _fit_and_score
estimator.fit(X_train, y_train, **fit_params)
File "c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\base.py", line 1144, in wrapper
estimator._validate_params()
File "c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\base.py", line 637, in _validate_params
validate_parameter_constraints(
File "c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\utils\_param_validation.py", line 95, in validate_parameter_constraints
raise InvalidParameterError(
sklearn.utils._param_validation.InvalidParameterError: The 'estimator' parameter of CalibratedClassifierCV must be an object implementing 'fit' and 'predict_proba', an object implementing 'fit' and 'decision_function' or None. Got 'estimator instance' instead.
c:\Users\DELL-G5\AppData\Local\Programs\Python\Python310\lib\site-packages\sklearn\model_selection\_search.py:976: UserWarning:
One or more of the test scores are non-finite: [ nan nan nan nan nan nan
0.49701389 0.49701389 0.49701389 0.50447917 0.50447917 0.50447917]
Out[ ]:
GridSearchCV(cv=3, estimator=CalibratedClassifierCV(),
param_grid=[{'estimator': ['estimator instance', None],
'method': ['sigmoid', 'isotonic'],
'n_jobs': [-1, 10, 16]}],
scoring='accuracy')In a Jupyter environment, please rerun this cell to show the HTML representation or trust the notebook. On GitHub, the HTML representation is unable to render, please try loading this page with nbviewer.org.
GridSearchCV(cv=3, estimator=CalibratedClassifierCV(),
param_grid=[{'estimator': ['estimator instance', None],
'method': ['sigmoid', 'isotonic'],
'n_jobs': [-1, 10, 16]}],
scoring='accuracy')CalibratedClassifierCV()
CalibratedClassifierCV()
The accuracy, precision, recall, and f1 scores, for the test set :
In [ ]:
P1=grid_search.predict(train)
P2=grid_search.predict(test)
print("Best parameter = "+str(grid_search.best_params_))
print("Best estimator = "+str(grid_search.best_estimator_))
print('The accuracy for CalibratedClassifierCV = '+str(grid_search.best_score_))
print("precision = "+str(precision_score(l2,P2)))
print("recall = "+str(recall_score(l2,P2)))
print("f1_score = "+str(f1_score(l2,P2)))
Best parameter = {'estimator': None, 'method': 'isotonic', 'n_jobs': -1}
Best estimator = CalibratedClassifierCV(method='isotonic', n_jobs=-1)
The accuracy for CalibratedClassifierCV = 0.5044791666666667
precision = 0.5051597051597052
recall = 0.5741413013124825
f1_score = 0.5374460854790223
The confusion matrix for:
1-training set
2-testing setIn [ ]:
ConfusionMatrixDisplay.from_predictions(l1,P1)
ConfusionMatrixDisplay.from_predictions(l2,P2)
plt.show()
SGDClassifier used to find the tradeoff between the precision and the recall¶
In [ ]:
from sklearn.linear_model import SGDClassifier
sgd_clf = SGDClassifier(random_state=42)
sgd_clf.fit(train, l1)
Out[ ]:
SGDClassifier(random_state=42)In a Jupyter environment, please rerun this cell to show the HTML representation or trust the notebook.
On GitHub, the HTML representation is unable to render, please try loading this page with nbviewer.org.
SGDClassifier(random_state=42)
The accuracy, precision, recall, and f1 scores, for the test set :
In [ ]:
P1=cross_val_predict(sgd_clf, train, l1, cv=3)
P2=cross_val_predict(sgd_clf, test, l2, cv=3)
print('The accuracy for SGDClassifier = '+str(accuracy_score(l2,P2)))
print("precision = "+str(precision_score(l2,P2)))
print("recall = "+str(recall_score(l2,P2)))
print("f1_score = "+str(f1_score(l2,P2)))
The accuracy for SGDClassifier = 0.4990277777777778 precision = 0.49711367673179396 recall = 0.6252443451549846 f1_score = 0.553865182436611
In [ ]:
l2_scores = cross_val_predict(sgd_clf, test, l2, cv=3,
method="decision_function")
precisions, recalls, thresholds = precision_recall_curve(l2, l2_scores)
print("Precisions : ",precisions,"\n\n","Recalls : ", recalls,"\n\n", "Thresholds : ",thresholds)
Precisions : [0.49736111 0.4974302 0.49749931 ... 0.5 0. 1. ] Recalls : [1.00000000e+00 1.00000000e+00 1.00000000e+00 ... 2.79251606e-04 0.00000000e+00 0.00000000e+00] Thresholds : [-3.19028312 -2.97208326 -2.8647555 ... 3.23812622 3.29804328 3.30090686]
In [ ]:
threshold=0.7
plt.figure(figsize=(8, 4))
plt.plot(thresholds, precisions[:-1], "k--", label="Precision", linewidth=2)
plt.plot(thresholds, recalls[:-1], "c-", label="Recall", linewidth=2)
plt.vlines(threshold, 0, 1.0, "k", "dotted", label="threshold")
idx = (thresholds >= threshold).argmax()
plt.plot(thresholds[idx], precisions[idx], "ko")
plt.plot(thresholds[idx], recalls[idx], "co")
plt.axis([-4,4, 0, 1.1])
plt.grid()
plt.xlabel("Threshold")
plt.legend(loc="upper right")
plt.show()
PART 3¶
Neural networks
Three models are created in this part, two of which are sequential, and one is functional, as shown in the table below :
| Properties | seqModel1 | seqModel2 | funcModel |
|---|---|---|---|
| Activation function | RelU | elU | RelU |
| Optimizer | Adam | RMSProp | SGD |
| Initializer | he_uniform | he_uniform | he_uniform |
| Dropout | Yes (0.15) | Yes (0.15) | No |
Here we split the training set to get validation set which represents 20% of the training set¶
In [ ]:
x_train= train[:23040]
y_train = l1[:23040]
x_valid= train[23040:]
y_valid = l1[23040:]
x_train.shape
Out[ ]:
(23040, 16)
seqModel1 :¶
In [ ]:
tf.random.set_seed(42)
model = tf.keras.Sequential()
model.add(layers.InputLayer(input_shape=(16)))
model.add(layers.Dense(400,kernel_initializer="he_normal",activation="relu"))
model.add(layers.Dropout(0.15))
model.add(layers.BatchNormalization())
model.add(layers.Dense(200,kernel_initializer="he_normal",activation="relu"))
model.add(layers.Dropout(0.15))
model.add(layers.BatchNormalization())
model.add(layers.Dense(100,kernel_initializer="he_normal",activation="relu"))
model.add(layers.Dropout(0.15))
model.add(layers.BatchNormalization())
model.add(layers.Dense(50,kernel_initializer="he_normal",activation="relu"))
model.add(layers.Dropout(0.15))
model.add(layers.BatchNormalization())
model.add(layers.Dense(25,kernel_initializer="he_normal",activation="relu"))
model.add(layers.Dropout(0.15))
model.add(layers.BatchNormalization())
model.add(layers.Dense(2,kernel_initializer="he_normal",activation="softmax"))
In [ ]:
model.summary()
Model: "sequential"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
dense (Dense) (None, 400) 6800
dropout (Dropout) (None, 400) 0
batch_normalization (Batch (None, 400) 1600
Normalization)
dense_1 (Dense) (None, 200) 80200
dropout_1 (Dropout) (None, 200) 0
batch_normalization_1 (Bat (None, 200) 800
chNormalization)
dense_2 (Dense) (None, 100) 20100
dropout_2 (Dropout) (None, 100) 0
batch_normalization_2 (Bat (None, 100) 400
chNormalization)
dense_3 (Dense) (None, 50) 5050
dropout_3 (Dropout) (None, 50) 0
batch_normalization_3 (Bat (None, 50) 200
chNormalization)
dense_4 (Dense) (None, 25) 1275
dropout_4 (Dropout) (None, 25) 0
batch_normalization_4 (Bat (None, 25) 100
chNormalization)
dense_5 (Dense) (None, 2) 52
=================================================================
Total params: 116577 (455.38 KB)
Trainable params: 115027 (449.32 KB)
Non-trainable params: 1550 (6.05 KB)
_________________________________________________________________
In [ ]:
tf.keras.utils.plot_model(model, "medstu_model0.png", show_shapes=True)
Out[ ]:
In [ ]:
optimizer = keras.optimizers.Adam(learning_rate=1e-3)
model.compile(loss="sparse_categorical_crossentropy",
optimizer=optimizer , metrics=["accuracy"])
In [ ]:
early_stopping_cb = tf.keras.callbacks.EarlyStopping(monitor="val_loss",patience=20)
out = model.fit(x_train, y_train, epochs=60, validation_data=(x_valid, y_valid),callbacks=[early_stopping_cb])
Epoch 1/60 720/720 [==============================] - 4s 3ms/step - loss: 0.7910 - accuracy: 0.4990 - val_loss: 0.6982 - val_accuracy: 0.4979 Epoch 2/60 720/720 [==============================] - 2s 3ms/step - loss: 0.7065 - accuracy: 0.5023 - val_loss: 0.6943 - val_accuracy: 0.5023 Epoch 3/60 720/720 [==============================] - 2s 3ms/step - loss: 0.6967 - accuracy: 0.5087 - val_loss: 0.6946 - val_accuracy: 0.4957 Epoch 4/60 720/720 [==============================] - 2s 3ms/step - loss: 0.6952 - accuracy: 0.5030 - val_loss: 0.6931 - val_accuracy: 0.5097 Epoch 5/60 720/720 [==============================] - 2s 3ms/step - loss: 0.6942 - accuracy: 0.5044 - val_loss: 0.6934 - val_accuracy: 0.5016 Epoch 6/60 720/720 [==============================] - 2s 3ms/step - loss: 0.6934 - accuracy: 0.5126 - val_loss: 0.6935 - val_accuracy: 0.5104 Epoch 7/60 720/720 [==============================] - 2s 3ms/step - loss: 0.6938 - accuracy: 0.5072 - val_loss: 0.6936 - val_accuracy: 0.5007 Epoch 8/60 720/720 [==============================] - 2s 3ms/step - loss: 0.6934 - accuracy: 0.5075 - val_loss: 0.6936 - val_accuracy: 0.4920 Epoch 9/60 720/720 [==============================] - 2s 3ms/step - loss: 0.6931 - accuracy: 0.5151 - val_loss: 0.6935 - val_accuracy: 0.5021 Epoch 10/60 720/720 [==============================] - 2s 3ms/step - loss: 0.6930 - accuracy: 0.5122 - val_loss: 0.6938 - val_accuracy: 0.5002 Epoch 11/60 720/720 [==============================] - 2s 3ms/step - loss: 0.6929 - accuracy: 0.5157 - val_loss: 0.6938 - val_accuracy: 0.5042 Epoch 12/60 720/720 [==============================] - 2s 3ms/step - loss: 0.6925 - accuracy: 0.5171 - val_loss: 0.6943 - val_accuracy: 0.5036 Epoch 13/60 720/720 [==============================] - 2s 3ms/step - loss: 0.6924 - accuracy: 0.5139 - val_loss: 0.6944 - val_accuracy: 0.5047 Epoch 14/60 720/720 [==============================] - 2s 3ms/step - loss: 0.6919 - accuracy: 0.5250 - val_loss: 0.6959 - val_accuracy: 0.5059 Epoch 15/60 720/720 [==============================] - 2s 3ms/step - loss: 0.6918 - accuracy: 0.5229 - val_loss: 0.6945 - val_accuracy: 0.5026 Epoch 16/60 720/720 [==============================] - 2s 3ms/step - loss: 0.6916 - accuracy: 0.5262 - val_loss: 0.6953 - val_accuracy: 0.5056 Epoch 17/60 720/720 [==============================] - 2s 3ms/step - loss: 0.6905 - accuracy: 0.5289 - val_loss: 0.6957 - val_accuracy: 0.5071 Epoch 18/60 720/720 [==============================] - 2s 3ms/step - loss: 0.6906 - accuracy: 0.5299 - val_loss: 0.6946 - val_accuracy: 0.5106 Epoch 19/60 720/720 [==============================] - 2s 3ms/step - loss: 0.6891 - accuracy: 0.5349 - val_loss: 0.6979 - val_accuracy: 0.4967 Epoch 20/60 720/720 [==============================] - 2s 3ms/step - loss: 0.6891 - accuracy: 0.5386 - val_loss: 0.6974 - val_accuracy: 0.5068 Epoch 21/60 720/720 [==============================] - 2s 3ms/step - loss: 0.6872 - accuracy: 0.5457 - val_loss: 0.6955 - val_accuracy: 0.5078 Epoch 22/60 720/720 [==============================] - 2s 3ms/step - loss: 0.6864 - accuracy: 0.5486 - val_loss: 0.6984 - val_accuracy: 0.5106 Epoch 23/60 720/720 [==============================] - 2s 3ms/step - loss: 0.6860 - accuracy: 0.5518 - val_loss: 0.6963 - val_accuracy: 0.5052 Epoch 24/60 720/720 [==============================] - 2s 3ms/step - loss: 0.6847 - accuracy: 0.5540 - val_loss: 0.6977 - val_accuracy: 0.5135
In [ ]:
out.params
Out[ ]:
{'verbose': 1, 'epochs': 60, 'steps': 720}
In [ ]:
t_loss,t_accuracy=model.evaluate(test, l2)
1/225 [..............................] - ETA: 4s - loss: 0.7606 - accuracy: 0.2812225/225 [==============================] - 0s 1ms/step - loss: 0.6960 - accuracy: 0.5113
In [ ]:
model.evaluate(test,l2)
X = test[:20]
y_proba = model.predict(X)
y_proba.round(2)
y_pred = y_proba.argmax(axis=-1)
print("predictions: "+str(y_pred))
x=l2[:20].tolist()
print("labels: "+str(x))
225/225 [==============================] - 0s 1ms/step - loss: 0.6960 - accuracy: 0.5113 1/1 [==============================] - 0s 143ms/step predictions: [1 0 0 1 0 0 1 1 1 0 1 1 0 0 0 0 1 1 1 1] labels: [0, 0, 1, 0, 1, 0, 0, 0, 0, 1, 1, 0, 1, 1, 1, 0, 0, 1, 1, 0]
In [ ]:
# "Accuracy"
plt.plot(out.history['accuracy'])
plt.plot(out.history['val_accuracy'])
plt.title('model accuracy')
plt.ylabel('accuracy')
plt.xlabel('epoch')
plt.legend(['train', 'validation'], loc='upper left')
plt.show()
# "Loss"
plt.plot(out.history['loss'])
plt.plot(out.history['val_loss'])
plt.title('model loss')
plt.ylabel('loss')
plt.xlabel('epoch')
plt.legend(['train', 'validation'], loc='upper left')
plt.show()
In [ ]:
list1=["train_data",'test_data','valid_data']
list2=[max(out.history['accuracy']),t_accuracy,max(out.history['val_accuracy'])]
plt.bar(list1,list2,color=['plum','lavender','pink'])
plt.title('accuracy for validation, training and testing data')
plt.ylabel('accuracy')
Out[ ]:
Text(0, 0.5, 'accuracy')
seqModel2 :¶
In [ ]:
tf.random.set_seed(42)
model = tf.keras.Sequential()
model.add(layers.InputLayer(input_shape=(16)))
model.add(layers.Dense(400,kernel_initializer="he_normal"))
model.add(layers.BatchNormalization())
model.add(layers.Activation("elu"))
model.add(layers.Dropout(0.15))
model.add(layers.Dense(200,kernel_initializer="he_normal"))
model.add(layers.BatchNormalization())
model.add(layers.Activation("elu"))
model.add(layers.Dropout(0.15))
model.add(layers.Dense(100,kernel_initializer="he_normal"))
model.add(layers.BatchNormalization())
model.add(layers.Activation("elu"))
model.add(layers.Dropout(0.15))
model.add(layers.Dense(50,kernel_initializer="he_normal"))
model.add(layers.BatchNormalization())
model.add(layers.Activation("elu"))
model.add(layers.Dropout(0.15))
model.add(layers.Dense(25,kernel_initializer="he_normal"))
model.add(layers.BatchNormalization())
model.add(layers.Activation("elu"))
model.add(layers.Dropout(0.15))
model.add(layers.Dense(2,kernel_initializer="he_normal",activation="softmax"))
In [ ]:
model.summary()
Model: "sequential_1"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
dense_6 (Dense) (None, 400) 6800
batch_normalization_5 (Bat (None, 400) 1600
chNormalization)
activation (Activation) (None, 400) 0
dropout_5 (Dropout) (None, 400) 0
dense_7 (Dense) (None, 200) 80200
batch_normalization_6 (Bat (None, 200) 800
chNormalization)
activation_1 (Activation) (None, 200) 0
dropout_6 (Dropout) (None, 200) 0
dense_8 (Dense) (None, 100) 20100
batch_normalization_7 (Bat (None, 100) 400
chNormalization)
activation_2 (Activation) (None, 100) 0
dropout_7 (Dropout) (None, 100) 0
dense_9 (Dense) (None, 50) 5050
batch_normalization_8 (Bat (None, 50) 200
chNormalization)
activation_3 (Activation) (None, 50) 0
dropout_8 (Dropout) (None, 50) 0
dense_10 (Dense) (None, 25) 1275
batch_normalization_9 (Bat (None, 25) 100
chNormalization)
activation_4 (Activation) (None, 25) 0
dropout_9 (Dropout) (None, 25) 0
dense_11 (Dense) (None, 2) 52
=================================================================
Total params: 116577 (455.38 KB)
Trainable params: 115027 (449.32 KB)
Non-trainable params: 1550 (6.05 KB)
_________________________________________________________________
In [ ]:
optimizer = keras.optimizers.RMSprop(learning_rate=1e-3)
model.compile(loss="sparse_categorical_crossentropy",
optimizer=optimizer,
metrics=["accuracy"])
In [ ]:
tf.keras.utils.plot_model(model, "medstu_model1.png", show_shapes=True)
Out[ ]:
In [ ]:
early_stopping_cb = tf.keras.callbacks.EarlyStopping(monitor="val_loss",patience=20)
out = model.fit(x_train, y_train, epochs=60, validation_data=(x_valid, y_valid),callbacks=[early_stopping_cb])
Epoch 1/60 720/720 [==============================] - 4s 3ms/step - loss: 0.7313 - accuracy: 0.5009 - val_loss: 0.6973 - val_accuracy: 0.5012 Epoch 2/60 720/720 [==============================] - 2s 3ms/step - loss: 0.7024 - accuracy: 0.5070 - val_loss: 0.6940 - val_accuracy: 0.4995 Epoch 3/60 720/720 [==============================] - 2s 3ms/step - loss: 0.6973 - accuracy: 0.5046 - val_loss: 0.6945 - val_accuracy: 0.5038 Epoch 4/60 720/720 [==============================] - 2s 3ms/step - loss: 0.6953 - accuracy: 0.5002 - val_loss: 0.6934 - val_accuracy: 0.5057 Epoch 5/60 720/720 [==============================] - 2s 3ms/step - loss: 0.6945 - accuracy: 0.4973 - val_loss: 0.6941 - val_accuracy: 0.4979 Epoch 6/60 720/720 [==============================] - 2s 3ms/step - loss: 0.6938 - accuracy: 0.5062 - val_loss: 0.6933 - val_accuracy: 0.5019 Epoch 7/60 720/720 [==============================] - 2s 3ms/step - loss: 0.6938 - accuracy: 0.5042 - val_loss: 0.6933 - val_accuracy: 0.5134 Epoch 8/60 720/720 [==============================] - 2s 3ms/step - loss: 0.6937 - accuracy: 0.5050 - val_loss: 0.6929 - val_accuracy: 0.5113 Epoch 9/60 720/720 [==============================] - 2s 3ms/step - loss: 0.6933 - accuracy: 0.5083 - val_loss: 0.6927 - val_accuracy: 0.5151 Epoch 10/60 720/720 [==============================] - 2s 3ms/step - loss: 0.6936 - accuracy: 0.5086 - val_loss: 0.6936 - val_accuracy: 0.4908 Epoch 11/60 720/720 [==============================] - 2s 3ms/step - loss: 0.6930 - accuracy: 0.5103 - val_loss: 0.6931 - val_accuracy: 0.5101 Epoch 12/60 720/720 [==============================] - 2s 3ms/step - loss: 0.6932 - accuracy: 0.5106 - val_loss: 0.6936 - val_accuracy: 0.5075 Epoch 13/60 720/720 [==============================] - 2s 3ms/step - loss: 0.6932 - accuracy: 0.5059 - val_loss: 0.6940 - val_accuracy: 0.5066 Epoch 14/60 720/720 [==============================] - 2s 3ms/step - loss: 0.6932 - accuracy: 0.5094 - val_loss: 0.6934 - val_accuracy: 0.5125 Epoch 15/60 720/720 [==============================] - 2s 3ms/step - loss: 0.6931 - accuracy: 0.5090 - val_loss: 0.6938 - val_accuracy: 0.5042 Epoch 16/60 720/720 [==============================] - 2s 3ms/step - loss: 0.6930 - accuracy: 0.5115 - val_loss: 0.6939 - val_accuracy: 0.5038 Epoch 17/60 720/720 [==============================] - 2s 3ms/step - loss: 0.6928 - accuracy: 0.5118 - val_loss: 0.6958 - val_accuracy: 0.5167 Epoch 18/60 720/720 [==============================] - 2s 3ms/step - loss: 0.6931 - accuracy: 0.5070 - val_loss: 0.6947 - val_accuracy: 0.4997 Epoch 19/60 720/720 [==============================] - 2s 3ms/step - loss: 0.6928 - accuracy: 0.5130 - val_loss: 0.6948 - val_accuracy: 0.5014 Epoch 20/60 720/720 [==============================] - 2s 3ms/step - loss: 0.6930 - accuracy: 0.5094 - val_loss: 0.6937 - val_accuracy: 0.5089 Epoch 21/60 720/720 [==============================] - 2s 3ms/step - loss: 0.6922 - accuracy: 0.5173 - val_loss: 0.6937 - val_accuracy: 0.5042 Epoch 22/60 720/720 [==============================] - 2s 3ms/step - loss: 0.6926 - accuracy: 0.5159 - val_loss: 0.6948 - val_accuracy: 0.5071 Epoch 23/60 720/720 [==============================] - 2s 3ms/step - loss: 0.6926 - accuracy: 0.5150 - val_loss: 0.6942 - val_accuracy: 0.5122 Epoch 24/60 720/720 [==============================] - 2s 3ms/step - loss: 0.6921 - accuracy: 0.5185 - val_loss: 0.6937 - val_accuracy: 0.5104 Epoch 25/60 720/720 [==============================] - 2s 3ms/step - loss: 0.6924 - accuracy: 0.5198 - val_loss: 0.6938 - val_accuracy: 0.5101 Epoch 26/60 720/720 [==============================] - 2s 3ms/step - loss: 0.6921 - accuracy: 0.5222 - val_loss: 0.6936 - val_accuracy: 0.5130 Epoch 27/60 720/720 [==============================] - 2s 3ms/step - loss: 0.6921 - accuracy: 0.5201 - val_loss: 0.6939 - val_accuracy: 0.5052 Epoch 28/60 720/720 [==============================] - 2s 3ms/step - loss: 0.6918 - accuracy: 0.5192 - val_loss: 0.6954 - val_accuracy: 0.5101 Epoch 29/60 720/720 [==============================] - 2s 3ms/step - loss: 0.6923 - accuracy: 0.5165 - val_loss: 0.6946 - val_accuracy: 0.5019
In [ ]:
t_val,t_accuracy=model.evaluate(test, l2)
1/225 [..............................] - ETA: 4s - loss: 0.7088 - accuracy: 0.3125225/225 [==============================] - 0s 1ms/step - loss: 0.6934 - accuracy: 0.5106
In [ ]:
model.evaluate(test,l2)
X = test[:20]
y_proba = model.predict(X)
y_proba.round(2)
y_pred = y_proba.argmax(axis=-1)
print("predictions: "+str(y_pred))
x=l2[:20].tolist()
print("labels: "+str(x))
225/225 [==============================] - 0s 1ms/step - loss: 0.6934 - accuracy: 0.5106 1/1 [==============================] - 0s 116ms/step predictions: [0 0 0 0 1 0 1 1 1 0 0 1 0 0 0 0 1 0 0 0] labels: [0, 0, 1, 0, 1, 0, 0, 0, 0, 1, 1, 0, 1, 1, 1, 0, 0, 1, 1, 0]
In [ ]:
print(out.history.keys())
# "Accuracy"
plt.plot(out.history['accuracy'])
plt.plot(out.history['val_accuracy'])
plt.title('model accuracy')
plt.ylabel('accuracy')
plt.xlabel('epoch')
plt.legend(['train', 'validation'], loc='upper left')
plt.grid(True)
plt.show()
# "Loss"
plt.plot(out.history['loss'])
plt.plot(out.history['val_loss'])
plt.title('model loss')
plt.ylabel('loss')
plt.xlabel('epoch')
plt.legend(['train', 'validation'], loc='upper right')
plt.grid(True)
plt.show()
dict_keys(['loss', 'accuracy', 'val_loss', 'val_accuracy'])
In [ ]:
list1=["train_data",'test_data','valid_data']
list2=[max(out.history['accuracy']),t_accuracy,max(out.history['val_accuracy'])]
plt.bar(list1,list2,color=['plum','lavender','pink'])
plt.title('accuracy for validation training and testing data')
plt.ylabel('accuracy')
Out[ ]:
Text(0, 0.5, 'accuracy')
funcModel :¶
In [ ]:
L1=layers.Input(shape=[6])
L2=layers.Input(shape=[12])
norm_1 = layers.BatchNormalization()(L1)
hidden1 =layers.Dense(50, activation="relu",kernel_initializer="he_uniform")(norm_1)
norm_2 = layers.BatchNormalization()(hidden1)
hidden2 =layers.Dense(50, activation="relu",kernel_initializer="he_uniform")(norm_2)
norm_3 = layers.BatchNormalization()(hidden2)
norm_4 = layers.BatchNormalization()(L2)
concat =layers.concatenate([norm_4, norm_3])
output1 = keras.layers.Dense(1 , name='Moutput')(concat)
output2 = keras.layers.Dense(1 , name='Aoutput')(norm_3)
model = keras.models.Model(inputs=[L1,L2], outputs=[output1,output2])
In [ ]:
model.summary()
Model: "model_1"
__________________________________________________________________________________________________
Layer (type) Output Shape Param # Connected to
==================================================================================================
input_5 (InputLayer) [(None, 6)] 0 []
batch_normalization_13 (Ba (None, 6) 24 ['input_5[0][0]']
tchNormalization)
dense_14 (Dense) (None, 50) 350 ['batch_normalization_13[0][0]
']
batch_normalization_14 (Ba (None, 50) 200 ['dense_14[0][0]']
tchNormalization)
input_6 (InputLayer) [(None, 12)] 0 []
dense_15 (Dense) (None, 50) 2550 ['batch_normalization_14[0][0]
']
batch_normalization_16 (Ba (None, 12) 48 ['input_6[0][0]']
tchNormalization)
batch_normalization_15 (Ba (None, 50) 200 ['dense_15[0][0]']
tchNormalization)
concatenate_1 (Concatenate (None, 62) 0 ['batch_normalization_16[0][0]
) ',
'batch_normalization_15[0][0]
']
Moutput (Dense) (None, 1) 63 ['concatenate_1[0][0]']
Aoutput (Dense) (None, 1) 51 ['batch_normalization_15[0][0]
']
==================================================================================================
Total params: 3486 (13.62 KB)
Trainable params: 3250 (12.70 KB)
Non-trainable params: 236 (944.00 Byte)
__________________________________________________________________________________________________
In [ ]:
tf.keras.utils.plot_model(model, "medstu_model2.png", show_shapes=True)
Out[ ]:
In [ ]:
model.compile(loss=["mean_squared_error","mean_squared_error"], optimizer=keras.optimizers.SGD(learning_rate=1e-3), metrics=["accuracy"])
In [ ]:
X_train_A, X_train_B = x_train.iloc[:, :6], x_train.iloc[:, 4:]
X_valid_A, X_valid_B = x_valid.iloc[:, :6], x_valid.iloc[:, 4:]
X_test_A, X_test_B = test.iloc[:, :6], test.iloc[:, 4:]
X_new_A, X_new_B = X_test_A.iloc[:6], X_test_B.iloc[:6]
In [ ]:
history = model.fit((X_train_A, X_train_B), y_train, epochs=40,validation_data=((X_valid_A, X_valid_B), y_valid))
Epoch 1/40 720/720 [==============================] - 3s 2ms/step - loss: 1.0952 - Moutput_loss: 0.5533 - Aoutput_loss: 0.5419 - Moutput_accuracy: 0.5000 - Aoutput_accuracy: 0.4953 - val_loss: 0.6960 - val_Moutput_loss: 0.3438 - val_Aoutput_loss: 0.3522 - val_Moutput_accuracy: 0.4984 - val_Aoutput_accuracy: 0.4958 Epoch 2/40 720/720 [==============================] - 1s 2ms/step - loss: 0.6814 - Moutput_loss: 0.3334 - Aoutput_loss: 0.3479 - Moutput_accuracy: 0.5007 - Aoutput_accuracy: 0.4935 - val_loss: 0.6072 - val_Moutput_loss: 0.3008 - val_Aoutput_loss: 0.3064 - val_Moutput_accuracy: 0.4988 - val_Aoutput_accuracy: 0.5054 Epoch 3/40 720/720 [==============================] - 1s 2ms/step - loss: 0.6176 - Moutput_loss: 0.3041 - Aoutput_loss: 0.3135 - Moutput_accuracy: 0.5004 - Aoutput_accuracy: 0.4976 - val_loss: 0.5741 - val_Moutput_loss: 0.2851 - val_Aoutput_loss: 0.2890 - val_Moutput_accuracy: 0.4967 - val_Aoutput_accuracy: 0.5068 Epoch 4/40 720/720 [==============================] - 2s 2ms/step - loss: 0.5857 - Moutput_loss: 0.2896 - Aoutput_loss: 0.2961 - Moutput_accuracy: 0.5033 - Aoutput_accuracy: 0.5030 - val_loss: 0.5540 - val_Moutput_loss: 0.2756 - val_Aoutput_loss: 0.2785 - val_Moutput_accuracy: 0.5049 - val_Aoutput_accuracy: 0.5056 Epoch 5/40 720/720 [==============================] - 2s 2ms/step - loss: 0.5667 - Moutput_loss: 0.2803 - Aoutput_loss: 0.2864 - Moutput_accuracy: 0.5034 - Aoutput_accuracy: 0.4994 - val_loss: 0.5408 - val_Moutput_loss: 0.2693 - val_Aoutput_loss: 0.2715 - val_Moutput_accuracy: 0.5005 - val_Aoutput_accuracy: 0.5082 Epoch 6/40 720/720 [==============================] - 1s 2ms/step - loss: 0.5538 - Moutput_loss: 0.2750 - Aoutput_loss: 0.2788 - Moutput_accuracy: 0.5065 - Aoutput_accuracy: 0.5007 - val_loss: 0.5333 - val_Moutput_loss: 0.2657 - val_Aoutput_loss: 0.2677 - val_Moutput_accuracy: 0.5033 - val_Aoutput_accuracy: 0.5063 Epoch 7/40 720/720 [==============================] - 1s 2ms/step - loss: 0.5465 - Moutput_loss: 0.2717 - Aoutput_loss: 0.2748 - Moutput_accuracy: 0.5005 - Aoutput_accuracy: 0.5008 - val_loss: 0.5270 - val_Moutput_loss: 0.2628 - val_Aoutput_loss: 0.2642 - val_Moutput_accuracy: 0.5031 - val_Aoutput_accuracy: 0.5080 Epoch 8/40 720/720 [==============================] - 2s 3ms/step - loss: 0.5367 - Moutput_loss: 0.2666 - Aoutput_loss: 0.2701 - Moutput_accuracy: 0.5069 - Aoutput_accuracy: 0.5022 - val_loss: 0.5231 - val_Moutput_loss: 0.2608 - val_Aoutput_loss: 0.2623 - val_Moutput_accuracy: 0.5031 - val_Aoutput_accuracy: 0.5059 Epoch 9/40 720/720 [==============================] - 1s 2ms/step - loss: 0.5314 - Moutput_loss: 0.2644 - Aoutput_loss: 0.2670 - Moutput_accuracy: 0.5058 - Aoutput_accuracy: 0.5022 - val_loss: 0.5187 - val_Moutput_loss: 0.2590 - val_Aoutput_loss: 0.2597 - val_Moutput_accuracy: 0.5049 - val_Aoutput_accuracy: 0.5049 Epoch 10/40 720/720 [==============================] - 1s 2ms/step - loss: 0.5294 - Moutput_loss: 0.2643 - Aoutput_loss: 0.2652 - Moutput_accuracy: 0.4980 - Aoutput_accuracy: 0.4997 - val_loss: 0.5167 - val_Moutput_loss: 0.2579 - val_Aoutput_loss: 0.2588 - val_Moutput_accuracy: 0.5003 - val_Aoutput_accuracy: 0.5054 Epoch 11/40 720/720 [==============================] - 1s 2ms/step - loss: 0.5232 - Moutput_loss: 0.2607 - Aoutput_loss: 0.2625 - Moutput_accuracy: 0.5040 - Aoutput_accuracy: 0.5039 - val_loss: 0.5141 - val_Moutput_loss: 0.2565 - val_Aoutput_loss: 0.2576 - val_Moutput_accuracy: 0.5059 - val_Aoutput_accuracy: 0.5099 Epoch 12/40 720/720 [==============================] - 1s 2ms/step - loss: 0.5217 - Moutput_loss: 0.2599 - Aoutput_loss: 0.2618 - Moutput_accuracy: 0.5025 - Aoutput_accuracy: 0.5005 - val_loss: 0.5122 - val_Moutput_loss: 0.2559 - val_Aoutput_loss: 0.2563 - val_Moutput_accuracy: 0.5028 - val_Aoutput_accuracy: 0.5120 Epoch 13/40 720/720 [==============================] - 2s 3ms/step - loss: 0.5190 - Moutput_loss: 0.2590 - Aoutput_loss: 0.2599 - Moutput_accuracy: 0.5023 - Aoutput_accuracy: 0.5026 - val_loss: 0.5106 - val_Moutput_loss: 0.2550 - val_Aoutput_loss: 0.2556 - val_Moutput_accuracy: 0.5059 - val_Aoutput_accuracy: 0.5050 Epoch 14/40 720/720 [==============================] - 2s 2ms/step - loss: 0.5163 - Moutput_loss: 0.2577 - Aoutput_loss: 0.2586 - Moutput_accuracy: 0.5069 - Aoutput_accuracy: 0.5052 - val_loss: 0.5092 - val_Moutput_loss: 0.2543 - val_Aoutput_loss: 0.2549 - val_Moutput_accuracy: 0.5031 - val_Aoutput_accuracy: 0.5082 Epoch 15/40 720/720 [==============================] - 2s 2ms/step - loss: 0.5146 - Moutput_loss: 0.2566 - Aoutput_loss: 0.2580 - Moutput_accuracy: 0.5004 - Aoutput_accuracy: 0.5026 - val_loss: 0.5085 - val_Moutput_loss: 0.2540 - val_Aoutput_loss: 0.2545 - val_Moutput_accuracy: 0.5050 - val_Aoutput_accuracy: 0.5080 Epoch 16/40 720/720 [==============================] - 1s 2ms/step - loss: 0.5130 - Moutput_loss: 0.2558 - Aoutput_loss: 0.2571 - Moutput_accuracy: 0.5069 - Aoutput_accuracy: 0.5051 - val_loss: 0.5081 - val_Moutput_loss: 0.2538 - val_Aoutput_loss: 0.2543 - val_Moutput_accuracy: 0.4972 - val_Aoutput_accuracy: 0.5028 Epoch 17/40 720/720 [==============================] - 1s 2ms/step - loss: 0.5114 - Moutput_loss: 0.2553 - Aoutput_loss: 0.2561 - Moutput_accuracy: 0.5062 - Aoutput_accuracy: 0.5086 - val_loss: 0.5078 - val_Moutput_loss: 0.2539 - val_Aoutput_loss: 0.2539 - val_Moutput_accuracy: 0.5047 - val_Aoutput_accuracy: 0.5026 Epoch 18/40 720/720 [==============================] - 1s 2ms/step - loss: 0.5106 - Moutput_loss: 0.2550 - Aoutput_loss: 0.2556 - Moutput_accuracy: 0.5039 - Aoutput_accuracy: 0.5059 - val_loss: 0.5058 - val_Moutput_loss: 0.2527 - val_Aoutput_loss: 0.2532 - val_Moutput_accuracy: 0.5092 - val_Aoutput_accuracy: 0.5010 Epoch 19/40 720/720 [==============================] - 1s 2ms/step - loss: 0.5094 - Moutput_loss: 0.2544 - Aoutput_loss: 0.2551 - Moutput_accuracy: 0.5054 - Aoutput_accuracy: 0.5075 - val_loss: 0.5072 - val_Moutput_loss: 0.2535 - val_Aoutput_loss: 0.2536 - val_Moutput_accuracy: 0.5007 - val_Aoutput_accuracy: 0.4998 Epoch 20/40 720/720 [==============================] - 1s 2ms/step - loss: 0.5090 - Moutput_loss: 0.2545 - Aoutput_loss: 0.2545 - Moutput_accuracy: 0.5051 - Aoutput_accuracy: 0.5073 - val_loss: 0.5051 - val_Moutput_loss: 0.2524 - val_Aoutput_loss: 0.2527 - val_Moutput_accuracy: 0.5099 - val_Aoutput_accuracy: 0.5061 Epoch 21/40 720/720 [==============================] - 1s 2ms/step - loss: 0.5086 - Moutput_loss: 0.2539 - Aoutput_loss: 0.2547 - Moutput_accuracy: 0.5102 - Aoutput_accuracy: 0.5064 - val_loss: 0.5045 - val_Moutput_loss: 0.2521 - val_Aoutput_loss: 0.2524 - val_Moutput_accuracy: 0.5052 - val_Aoutput_accuracy: 0.5054 Epoch 22/40 720/720 [==============================] - 1s 1ms/step - loss: 0.5074 - Moutput_loss: 0.2534 - Aoutput_loss: 0.2540 - Moutput_accuracy: 0.5065 - Aoutput_accuracy: 0.5030 - val_loss: 0.5040 - val_Moutput_loss: 0.2518 - val_Aoutput_loss: 0.2522 - val_Moutput_accuracy: 0.5010 - val_Aoutput_accuracy: 0.5073 Epoch 23/40 720/720 [==============================] - 1s 2ms/step - loss: 0.5063 - Moutput_loss: 0.2529 - Aoutput_loss: 0.2534 - Moutput_accuracy: 0.5065 - Aoutput_accuracy: 0.5069 - val_loss: 0.5036 - val_Moutput_loss: 0.2516 - val_Aoutput_loss: 0.2520 - val_Moutput_accuracy: 0.4983 - val_Aoutput_accuracy: 0.5045 Epoch 24/40 720/720 [==============================] - 1s 1ms/step - loss: 0.5062 - Moutput_loss: 0.2529 - Aoutput_loss: 0.2533 - Moutput_accuracy: 0.5069 - Aoutput_accuracy: 0.5042 - val_loss: 0.5037 - val_Moutput_loss: 0.2517 - val_Aoutput_loss: 0.2520 - val_Moutput_accuracy: 0.5014 - val_Aoutput_accuracy: 0.5023 Epoch 25/40 720/720 [==============================] - 1s 1ms/step - loss: 0.5050 - Moutput_loss: 0.2521 - Aoutput_loss: 0.2528 - Moutput_accuracy: 0.5102 - Aoutput_accuracy: 0.5075 - val_loss: 0.5029 - val_Moutput_loss: 0.2513 - val_Aoutput_loss: 0.2516 - val_Moutput_accuracy: 0.5089 - val_Aoutput_accuracy: 0.5097 Epoch 26/40 720/720 [==============================] - 1s 1ms/step - loss: 0.5051 - Moutput_loss: 0.2526 - Aoutput_loss: 0.2525 - Moutput_accuracy: 0.5040 - Aoutput_accuracy: 0.5075 - val_loss: 0.5027 - val_Moutput_loss: 0.2511 - val_Aoutput_loss: 0.2516 - val_Moutput_accuracy: 0.5033 - val_Aoutput_accuracy: 0.5061 Epoch 27/40 720/720 [==============================] - 1s 1ms/step - loss: 0.5054 - Moutput_loss: 0.2526 - Aoutput_loss: 0.2528 - Moutput_accuracy: 0.5056 - Aoutput_accuracy: 0.5077 - val_loss: 0.5025 - val_Moutput_loss: 0.2511 - val_Aoutput_loss: 0.2514 - val_Moutput_accuracy: 0.5097 - val_Aoutput_accuracy: 0.4965 Epoch 28/40 720/720 [==============================] - 1s 1ms/step - loss: 0.5043 - Moutput_loss: 0.2519 - Aoutput_loss: 0.2524 - Moutput_accuracy: 0.5071 - Aoutput_accuracy: 0.5053 - val_loss: 0.5023 - val_Moutput_loss: 0.2510 - val_Aoutput_loss: 0.2513 - val_Moutput_accuracy: 0.5115 - val_Aoutput_accuracy: 0.4988 Epoch 29/40 720/720 [==============================] - 1s 1ms/step - loss: 0.5035 - Moutput_loss: 0.2516 - Aoutput_loss: 0.2519 - Moutput_accuracy: 0.5045 - Aoutput_accuracy: 0.5120 - val_loss: 0.5018 - val_Moutput_loss: 0.2507 - val_Aoutput_loss: 0.2511 - val_Moutput_accuracy: 0.5054 - val_Aoutput_accuracy: 0.5047 Epoch 30/40 720/720 [==============================] - 1s 1ms/step - loss: 0.5035 - Moutput_loss: 0.2517 - Aoutput_loss: 0.2518 - Moutput_accuracy: 0.5055 - Aoutput_accuracy: 0.5052 - val_loss: 0.5021 - val_Moutput_loss: 0.2509 - val_Aoutput_loss: 0.2512 - val_Moutput_accuracy: 0.5075 - val_Aoutput_accuracy: 0.5095 Epoch 31/40 720/720 [==============================] - 1s 1ms/step - loss: 0.5033 - Moutput_loss: 0.2515 - Aoutput_loss: 0.2517 - Moutput_accuracy: 0.5038 - Aoutput_accuracy: 0.5090 - val_loss: 0.5023 - val_Moutput_loss: 0.2510 - val_Aoutput_loss: 0.2512 - val_Moutput_accuracy: 0.5061 - val_Aoutput_accuracy: 0.5043 Epoch 32/40 720/720 [==============================] - 1s 2ms/step - loss: 0.5032 - Moutput_loss: 0.2514 - Aoutput_loss: 0.2518 - Moutput_accuracy: 0.5082 - Aoutput_accuracy: 0.5088 - val_loss: 0.5013 - val_Moutput_loss: 0.2505 - val_Aoutput_loss: 0.2508 - val_Moutput_accuracy: 0.5113 - val_Aoutput_accuracy: 0.5038 Epoch 33/40 720/720 [==============================] - 1s 2ms/step - loss: 0.5022 - Moutput_loss: 0.2511 - Aoutput_loss: 0.2512 - Moutput_accuracy: 0.5104 - Aoutput_accuracy: 0.5117 - val_loss: 0.5020 - val_Moutput_loss: 0.2509 - val_Aoutput_loss: 0.2511 - val_Moutput_accuracy: 0.5076 - val_Aoutput_accuracy: 0.5003 Epoch 34/40 720/720 [==============================] - 1s 2ms/step - loss: 0.5019 - Moutput_loss: 0.2508 - Aoutput_loss: 0.2511 - Moutput_accuracy: 0.5140 - Aoutput_accuracy: 0.5128 - val_loss: 0.5021 - val_Moutput_loss: 0.2510 - val_Aoutput_loss: 0.2511 - val_Moutput_accuracy: 0.5036 - val_Aoutput_accuracy: 0.5009 Epoch 35/40 720/720 [==============================] - 1s 2ms/step - loss: 0.5021 - Moutput_loss: 0.2509 - Aoutput_loss: 0.2512 - Moutput_accuracy: 0.5093 - Aoutput_accuracy: 0.5094 - val_loss: 0.5016 - val_Moutput_loss: 0.2508 - val_Aoutput_loss: 0.2508 - val_Moutput_accuracy: 0.5066 - val_Aoutput_accuracy: 0.5047 Epoch 36/40 720/720 [==============================] - 1s 2ms/step - loss: 0.5022 - Moutput_loss: 0.2510 - Aoutput_loss: 0.2512 - Moutput_accuracy: 0.5081 - Aoutput_accuracy: 0.5122 - val_loss: 0.5022 - val_Moutput_loss: 0.2511 - val_Aoutput_loss: 0.2512 - val_Moutput_accuracy: 0.5085 - val_Aoutput_accuracy: 0.5036 Epoch 37/40 720/720 [==============================] - 1s 2ms/step - loss: 0.5025 - Moutput_loss: 0.2511 - Aoutput_loss: 0.2513 - Moutput_accuracy: 0.5109 - Aoutput_accuracy: 0.5079 - val_loss: 0.5021 - val_Moutput_loss: 0.2510 - val_Aoutput_loss: 0.2511 - val_Moutput_accuracy: 0.5064 - val_Aoutput_accuracy: 0.5031 Epoch 38/40 720/720 [==============================] - 1s 2ms/step - loss: 0.5023 - Moutput_loss: 0.2510 - Aoutput_loss: 0.2513 - Moutput_accuracy: 0.5091 - Aoutput_accuracy: 0.5081 - val_loss: 0.5014 - val_Moutput_loss: 0.2507 - val_Aoutput_loss: 0.2507 - val_Moutput_accuracy: 0.5056 - val_Aoutput_accuracy: 0.5078 Epoch 39/40 720/720 [==============================] - 2s 2ms/step - loss: 0.5018 - Moutput_loss: 0.2508 - Aoutput_loss: 0.2510 - Moutput_accuracy: 0.5116 - Aoutput_accuracy: 0.5102 - val_loss: 0.5014 - val_Moutput_loss: 0.2507 - val_Aoutput_loss: 0.2507 - val_Moutput_accuracy: 0.5085 - val_Aoutput_accuracy: 0.5068 Epoch 40/40 720/720 [==============================] - 1s 2ms/step - loss: 0.5014 - Moutput_loss: 0.2508 - Aoutput_loss: 0.2506 - Moutput_accuracy: 0.5068 - Aoutput_accuracy: 0.5128 - val_loss: 0.5011 - val_Moutput_loss: 0.2505 - val_Aoutput_loss: 0.2506 - val_Moutput_accuracy: 0.5054 - val_Aoutput_accuracy: 0.5005
In [ ]:
mse_test = model.evaluate((X_test_A, X_test_B), l2)
y_pred = model.predict((X_new_A, X_new_B))
225/225 [==============================] - 0s 970us/step - loss: 0.5019 - Moutput_loss: 0.2510 - Aoutput_loss: 0.2509 - Moutput_accuracy: 0.5049 - Aoutput_accuracy: 0.5058 1/1 [==============================] - 0s 103ms/step
In [ ]:
pd.DataFrame(history.history).plot(
figsize=(10, 7), ylim=[0,1], grid=True, xlabel="Epoch",
style=["r--", "r--.", "b-", "b-*"])
plt.legend(loc="upper right")
plt.show()
In [ ]: